A GLOBAL SOLUTION TO THE NONLINEAR MODEL-PREDICTIVE CONTROL ALGORITHMS USING POLYNOMIAL ARX MODELS

In nonlinear model predictive control algorithms, a nonlinear objectiv e function is minimized on-line at every sampling time. Finding a glob al optimum is not very easy as the objective function is generally non linear and nonconvex. In this paper we show that the structure of the polynomial ARX models lends model predictive algorithms some useful pr operties that are helpful in determining the global optimum solution. The approach used is based on transformation and change of variables t o recast the problem into a convex objective function with convex cons traints. A method is then proposed that guarantees a global solution t o the optimization problem. Copyright (C) 1996 Elsevier Science Ltd